HTML output for tbl

HTML output for tbl

[Joerg Sonnenberger]

Hi Kristaps et al,
looking at the output for queue(3), there is a rather obvious bug in the
width output. It repeats the width: xxx for every column, e.g. the third
column has three width in order. This magically works due to the
overwrite rules, but is far from pretty. I also wonder if we should use
min-width and screw IE 6. That and using a single table (not one per
row) would be enough to give queue(3) a properly formatted table without
having to go extra wide by switching spacing from ex to em.

Joerg
--
 To unsubscribe send an email to tech+unsubscribe@mdocml.bsd.lv

Re: HTML output for tbl

[Kristaps Dzonsons]

> looking at the output for queue(3), there is a rather obvious bug in the
> width output. It repeats the width: xxx for every column, e.g. the third
> column has three width in order. This magically works due to the
> overwrite rules, but is far from pretty. I also wonder if we should use
> min-width and screw IE 6. That and using a single table (not one per
> row) would be enough to give queue(3) a properly formatted table without
> having to go extra wide by switching spacing from ex to em.

Joerg,

This isn't really a bug: it's because tables can be interspersed with 
arbitrary macros, so each row should in theory be standalone.

However, I agree that this makes for ugly output, and having just tested 
it with SCALE_EM, I think it's best to keep a few bits of state and just 
reinitialise the table if broken up by other macros.  I'll write this up 
when I've a few minutes to myself.

And no, we can't screw IE6.

It just now occurs to me that, since CSS "cascades" atop the HTML, I can 
set <COL> pixel-widths as a safe default and let CSS, with its precise 
"em" widths, override these values.  This will make browsers without CSS 
also recognise mandoc -Thtml's tables, whether from tbl(7) or otherwise, 
which for now require the style-sheet for stipulating widths.

Thanks,

Kristaps
--
 To unsubscribe send an email to tech+unsubscribe@mdocml.bsd.lv

Re: HTML output for tbl

[Joerg Sonnenberger]

On Thu, Jan 13, 2011 at 01:36:43PM +0100, Kristaps Dzonsons wrote:
> >looking at the output for queue(3), there is a rather obvious bug in the
> >width output. It repeats the width: xxx for every column, e.g. the third
> >column has three width in order. This magically works due to the
> >overwrite rules, but is far from pretty. I also wonder if we should use
> >min-width and screw IE 6. That and using a single table (not one per
> >row) would be enough to give queue(3) a properly formatted table without
> >having to go extra wide by switching spacing from ex to em.
> 
> Joerg,
> 
> This isn't really a bug: it's because tables can be interspersed
> with arbitrary macros, so each row should in theory be standalone.
> 
> However, I agree that this makes for ugly output, and having just
> tested it with SCALE_EM, I think it's best to keep a few bits of
> state and just reinitialise the table if broken up by other macros.
> I'll write this up when I've a few minutes to myself.

ACK. I do prefer to have SCALE_EX as default, the width can go way too
large otherwise.

> And no, we can't screw IE6.

Well, the other option would be to require a bit JS for IE6. The point
is that it is the only larger browser that might still be used that
doesn't do min-width and also understands CSS. Worst case is that it
doesn't use the size hints -- I consider that acceptable as fallout for
getting much better output with useful browsers.

> It just now occurs to me that, since CSS "cascades" atop the HTML, I
> can set <COL> pixel-widths as a safe default and let CSS, with its
> precise "em" widths, override these values.  This will make browsers
> without CSS also recognise mandoc -Thtml's tables, whether from
> tbl(7) or otherwise, which for now require the style-sheet for
> stipulating widths.

I disagree somewhat. Using col is a good idea as it avoids redundant
markup. Trying to second guess the font width is prone to fail. That's
why specifying a minimal width is better -- it will still just stretch
if needed.

Joerg
--
 To unsubscribe send an email to tech+unsubscribe@mdocml.bsd.lv

Re: HTML output for tbl

[Kristaps Dzonsons]

>> However, I agree that this makes for ugly output, and having just
>> tested it with SCALE_EM, I think it's best to keep a few bits of
>> state and just reinitialise the table if broken up by other macros.
>> I'll write this up when I've a few minutes to myself.
>
> ACK. I do prefer to have SCALE_EX as default, the width can go way too
> large otherwise.

Joerg,

I just checked in a change to -Thtml for tables that has the following 
basic logic: when a new table is encountered, it sets column widths 
using <COL> and the "width" CSS style.

These columns are used until a macro is encountered.  At this point, the 
table is closed out and the macro is processed.

(This also works with block macros.)

Then when the first table-row is again encountered, the table is 
re-opened in the same way.

This allows for arbitrary breaks in the table flow; subsequent table 
rows, by being in the same TABLE, are all sized intelligently by the 
browser.

>> And no, we can't screw IE6.
>
> Well, the other option would be to require a bit JS for IE6. The point
> is that it is the only larger browser that might still be used that
> doesn't do min-width and also understands CSS. Worst case is that it
> doesn't use the size hints -- I consider that acceptable as fallout for
> getting much better output with useful browsers.
>
>> It just now occurs to me that, since CSS "cascades" atop the HTML, I
>> can set<COL>  pixel-widths as a safe default and let CSS, with its
>> precise "em" widths, override these values.  This will make browsers
>> without CSS also recognise mandoc -Thtml's tables, whether from
>> tbl(7) or otherwise, which for now require the style-sheet for
>> stipulating widths.
>
> I disagree somewhat. Using col is a good idea as it avoids redundant
> markup. Trying to second guess the font width is prone to fail. That's
> why specifying a minimal width is better -- it will still just stretch
> if needed.

Consider this.  If I hardcode <COL WIDTH="20" STYLE="min-width: 4ex;">, 
where "20" is (I'm just making these numbers up) a reasonable guess at 
the pixel-width of 4 x's, then "real" browsers can use the min-width and 
IE6, and any other non-CSS, will use the COL sizes for a reasonable spacing.

Thoughts?

Kristaps
--
 To unsubscribe send an email to tech+unsubscribe@mdocml.bsd.lv

Re: HTML output for tbl

[Joerg Sonnenberger]

On Thu, Jan 13, 2011 at 03:30:42PM +0100, Kristaps Dzonsons wrote:
> >>However, I agree that this makes for ugly output, and having just
> >>tested it with SCALE_EM, I think it's best to keep a few bits of
> >>state and just reinitialise the table if broken up by other macros.
> >>I'll write this up when I've a few minutes to myself.
> >
> >ACK. I do prefer to have SCALE_EX as default, the width can go way too
> >large otherwise.
> 
> Joerg,
> 
> I just checked in a change to -Thtml for tables that has the
> following basic logic: when a new table is encountered, it sets
> column widths using <COL> and the "width" CSS style.

This looks good.

> These columns are used until a macro is encountered.  At this point,
> the table is closed out and the macro is processed.
> 
> (This also works with block macros.)
> 
> Then when the first table-row is again encountered, the table is
> re-opened in the same way.
> 
> This allows for arbitrary breaks in the table flow; subsequent table
> rows, by being in the same TABLE, are all sized intelligently by the
> browser.

This sounds fine, too.

> >>And no, we can't screw IE6.
> >
> >Well, the other option would be to require a bit JS for IE6. The point
> >is that it is the only larger browser that might still be used that
> >doesn't do min-width and also understands CSS. Worst case is that it
> >doesn't use the size hints -- I consider that acceptable as fallout for
> >getting much better output with useful browsers.
> >
> >>It just now occurs to me that, since CSS "cascades" atop the HTML, I
> >>can set<COL>  pixel-widths as a safe default and let CSS, with its
> >>precise "em" widths, override these values.  This will make browsers
> >>without CSS also recognise mandoc -Thtml's tables, whether from
> >>tbl(7) or otherwise, which for now require the style-sheet for
> >>stipulating widths.
> >
> >I disagree somewhat. Using col is a good idea as it avoids redundant
> >markup. Trying to second guess the font width is prone to fail. That's
> >why specifying a minimal width is better -- it will still just stretch
> >if needed.
> 
> Consider this.  If I hardcode <COL WIDTH="20" STYLE="min-width:
> 4ex;">, where "20" is (I'm just making these numbers up) a
> reasonable guess at the pixel-width of 4 x's, then "real" browsers
> can use the min-width and IE6, and any other non-CSS, will use the
> COL sizes for a reasonable spacing.
> 
> Thoughts?

The question arrives if you have input that is too large to fit, but can
be broken apart like multiple words. For queue(3) the changed version
works because the "_FOREACH_REVERSE" is a single word. If you replace
the second _ with a space, it will be line wrapped. With min-width,
modern browser will try to increase the total width of the table to
avoid line breaks. Specifying both width and min-width is useless as
width overrides any min-width attribute.

Joerg
--
 To unsubscribe send an email to tech+unsubscribe@mdocml.bsd.lv